# Codebook for the PEMRA complaint gold set

Unit: one complaint as filed (one row of the regulator's monthly register). Code what the complainant objects to, in their words, not what you think was actually broadcast and not whether the complaint is justified.

Two things are coded for each complaint.

## A. Grounds (multi-label: tick every ground that applies, 0 or 1 in each column)

| Column | Ground | Tick when the complaint objects to... | PEMRA's own tick-boxes that map here |
|---|---|---|---|
| `g_religion` | Religion and sect | content against Islam or Islamic values; blasphemy or insult to religious figures, scholars or clerics; sectarian content; discrimination against or mockery of a religious community (Ahmadi, Hindu, Christian, Shia...); promotion of another religion | Sectarianism; Religious Discrimination; Detrimental to Islamic values |
| `g_hate` | Hate, violence and extremism | hate speech, incitement, threats, glorifying or encouraging violence, extremism; racial, ethnic or linguistic discrimination | Hatred; Encouraging Violence; Extremism; Racialism; Ethnic Discrimination |
| `g_morality` | Vulgarity and morality | indecency, obscenity, "bold" or intimate scenes, dress, dancing, item songs, sexual content in adverts, content "against our culture, values or family system", immorality in general | Vulgarity; Obscenity; Slander against basic cultural values, morality and good manners; Slander against moral life of the country |
| `g_gender` | Women and gender | demeaning or stereotyped portrayal of women (or men); harassment; mistreatment of female guests or victims on air; misogyny; complaints that a drama promotes "rebellious" wives or daughters; LGBTQ content complained of as gender content | Denigrates men or women... |
| `g_misinfo` | False, misleading or defamatory content | fake news, lies, unverified allegations, defamation of a named person or business, misreporting of an event | Slanders against individual or certain groups |
| `g_state` | State, nation and security | content against the armed forces, judiciary or other state institutions; "ideology of Pakistan"; national interest or security; contempt of court; "anti-state" narrative; leaking of security information | Detrimental to Ideology of Pakistan; Contempt of Court |
| `g_politics` | Political bias and anchor conduct | partisan coverage, propaganda for or against a party or leader, unbalanced panels, anchors or guests abusing politicians, political campaigns (TLP, PTI, PML-N...) described as being misrepresented | (none) |
| `g_foreign` | Foreign content | Indian songs, films, dramas or celebrities; Turkish or other foreign serials; foreign culture; foreign channels on cable | (none) |
| `g_other` | Other or none of the above | advertising volume or timing, cable or signal problems, channel availability, children's screen time in general, personal requests, unintelligible text. Tick ONLY if no other ground applies | |

Rules
- A complaint can carry several grounds. "This drama is vulgar and against Islam" is `g_morality` + `g_religion`. "Anchor lied about the army" is `g_misinfo` + `g_state`.
- Code the ground the complainant gives. A complaint that a drama is "against our religion and culture" gets religion + morality even if you think it is really about something else.
- Religion is ticked only when religion, a religious figure or a religious community is mentioned or clearly meant. "Against our values" alone is morality, not religion.
- Gender is ticked when the objection is about how women (or men) are shown or treated, including complaints that a drama shows women being disobedient, and complaints about how a female guest or victim was treated. A complaint that a drama shows a love affair is morality, not gender.
- Politics is for partisan or political-conduct complaints. A complaint that a channel defamed a politician by lying is misinfo + politics.
- Insulting a religious scholar (Maulana Tariq Jameel, a mufti) is religion. Insulting a politician is politics. Insulting a named private person is misinfo if the objection is to falsity, otherwise other.
- Campaign form letters are coded like any other complaint; the campaign is handled separately through the cluster weights.
- If the text is unintelligible (corrupted glyphs, a bare URL, a single word), tick `g_other` and put "unreadable" in notes.

## B. Target (multi-class: exactly one value in `target`)

| Value | Meaning |
|---|---|
| `news` | news bulletin, talk show, current affairs programme, anchor or reporter |
| `drama` | a drama serial or telefilm |
| `show` | morning show, game show, reality or comedy show, music programme, Ramadan transmission, any entertainment programme that is not a drama |
| `advert` | an advertisement or promo |
| `religious` | a religious programme or religious scholar's show |
| `channel` | the channel or cable operator as a whole (availability, signal, language, overall tone) with no particular programme |
| `unclear` | cannot tell |

## Columns in the sheet

`id` (complaint id), `year`, `month`, `channel` (as the register gives it; sometimes garbled), `lang`, `cluster_size` (how many near-identical complaints this one stands for), `text` (cleaned; Urdu letters are restored but sentence order inside Urdu passages can be scrambled by the PDF extraction), `gloss` (an English rendering of Urdu and Roman Urdu text, for speed; code from the original where they differ), then the nine `g_` columns, `target`, `notes`.

Please do not look at PEMRA's own ticks while coding (they are hidden in the sheet and only compared afterwards).